The system gain of the ultra-local model-free predictive control(ULM-MFPC)for dual active bridge(DAB)converters depends on physical parameters

and parameter mismatches such as inductance mismatch easily lead to control performance degradation.To address this problem

a data-driven model predictive control strategy based on a time-series model is proposed.First

an autoregressive with exogenous inputs(ARX)model containing only control input

output and random noise terms is established to accurately describe the system dynamics characteristics and reduce the dependence on accurate physical parameters.Then

a forgetting factor recursive least squares(FFRLS)is introduced to realize real-time adaptive update of model parameters

further enhancing the robustness of the model.Finally

a discrete prediction equation and a cost function are constructed

and control is achieved by solving the optimal phase-shift ratio.Simulation results show that the proposed strategy has better robustness than the traditional MFPC strategy

and under parameter mismatch conditions

it can effectively reduce output voltage ripple and eliminate the steady-state error.